HP 4395A Combination
Network, Spectrum, Impedance Analyzer

© Brooke Clarke, N6GCE 2001 - 2022

Introduction
Spectrum Analyzer
    Input Adapter
    Spectrum Monitoring
    Plots
    Frequency Converter
        Reverse Converter 120 to 240 VAC
        IEC 60906-1 to NEMA equipment power cord?
    Noise Source
Network Analyzer
    Test sets
    4380A
    Cal Kits   
    Couplers
    ZFDC-10-1 Coupler Frequency Response
    Measuring Cable Length
Impedance Analyzer
    43961A Impedance Test Set
Z Transform Impedance Method
    Background
    How it Works
    Calibration
    Z:Trans Fixtures
Other Impedance Info
Save State    
Signal Generator
Floppy Disk Drive
     Initialize Floppy
Display, Keyboard & Mouse
Time Gating
Applications
    Audio
    Amplifiers
    Antennas
    Crystal Oscillators
    Phase Noise
    Spectrum Plots
    VCO Testing
E5040A VXI Spectrum Analyzer
Reference Information
Firmware
Scalar Analyzers
Related
Links

---

Introduction

All of the IF signal processing is done by a DSP chip instead of using log amplifiers and analog filters as is done in older analog instruments.  This results in far superior performance than analog type boxes, especially in Spectrum Analyzer mode.  This is the replacement for the 4195 combination instrument and is newer than the 4396B which covers 100 kHz to 1.8 GHz box.  The 4395A covers 10 Hz to 500 MHz.

To get an feel for what the DSP method offers consider that an analog type spectrum analyzer takes 30 seconds to sweep a 100 KHz band using a  100 Hz resolution bandwidth.  The 4395A will make the same sweep in 0.3 seconds, i.e. 100 times faster.  The RBW filter widths offered on analog spectrum analyzers are much wider than the 1 Hz RBW offered on the 4395A.  This when combined with true I&Q detection allows measuring the true RMS value (an analog SA typically measures peak values) thus noise measurements can be displayed directly in /root hertz.

Spectrum Analyzer

10 Hz to 500 Mhz frequency range with a linear or log sweep.
If you have just the 4395A box you have a fully functional spectrum analyzer.  There is a hardware option 1D6 for time gating that will synchronize the sample time to an external pulse that's typically used for measuring things that rotate, like disk drives (E5022A).

The minimum RBW is 1 Hz.  This allows seeing very weak signals.  The narrowest IF bandwidth communications receiver I have is the NRD 545 with a DSP IF bandwidth of 10 Hz.  The 4395A has ten times lower bandwidth.  But the 4395A does not have any IF outputs.

Note that there is a way to display noise like signals as "per root Hz" and when this is done the display shows the units as /Hz because there is not character for SQRT.  In the Detection type select "sample" instead of positive or negative peak.  In this mode the DSP is computing SQRT(I*I + Q*Q), i.e. the actual power, unlike older analog spectrum analyzers that used peak detection.

The sweep speed is limited by the lowest displayed frequency.  When measuring audio frequencies the sweep needs to be slowed way down to get valid readings.

Input Adapter

To allow the audio output from a Squad Radio to be connected to the input of the spectrum analyzer an adapter is required to both block the DC signal (up to 6.x Volts DC is on top of the 1/2 Watt max audio signal to sense the connection of an external speaker and mute the internal speaker.) and attenuate the AC signal.  To do this and have a reasonably low cutoff frequency I used the following circuit:
Starting from the Spectrum analyzer BNC(m) connector on a Pomona model 2391 box that comes with BNC connectors already installed:
a 22 Ohm resistor to node 1
a 33 Ohm resistor from node 1 to ground
a 470 Ohm resistor from node 1 to node 2
the ground side of a 150 uF cap from node 2 to node 3
a 470 Ohm resistor from node 3 to node 4
a 33 Ohm resistor from node 4 to ground
a 22 Ohm resistor from node 4 to the center pin of the BNC(f) connector on the radio side of the adapter.

A Network analyzer response test shows the attenuation to be within a couple of tenths of a dB up to many tens of Mhz.
At the low frequency end at 21 Hz the response is low by 0.89 dB.  At 150 Hz the response is 0.1 dB low. This adapter has 40.0 dB flat attenuation over the useful audio spectrum.  That's not a problem since the 4395 has a huge dynamic range so loosing 40 dB is not an issue.

Spectrum Monitoring

Because of the 1 Hz RBW capability of the DSP filter it's possible to see very small signals, although it takes time to sweep wide bandwidths.  There is no IF output like on analog spectrum analyzers.  You can not connect an external demodulator.

By using the DISPLAY/DATA HOLD MAX function, the trace remembers the peaks to provide a visual record of all transmissions.  Example plot scan 30 to 90 MHz, 3 kHz RBW overnight.  3090A.TIF
118 - 136 MHz AM air band, used multiple passes of sub markers to get all the frequencies

Spectrum Plots

0 to 200 kHz
200 to 400 kHz
118 to 136 MHz
136 to 138 MHz
150 to 154 MHz
153 to 157 MHz 
Spectrum around WWVB 60 kHz - description of watching the 10 dB time code amplitude modulation

Frequency Converter

Got this from eBay with title: F2 2GHz Spectrum Analyzer Frequency Expander for AT5011
The Make & model number: Shenzhen Atten Electronics Ltd. AT5000-F2
Input: 2050 - 3050 MHz @ -27 -to -100 dBm.
Output: 0 to 1000 MHz

The problems are:

•  that they left out one of the transformer mounting screws, see photo above.  The unit came in a sealed plastic bag and there was no screw there.  I think the screw fell out before the unit was bagged.
  
• The transformer is for 220 VAC, not 120VAC.
  

Will be changing the transformer once the output specs are determined.  (nominal output voltage and current capacity).
The power supply board is fed with 3 wires from the transformer and feeds three wires to the RF converter.
There are a couple of 3-terminal regulator ICs:

• L7812CV
• L7805CV

I'm guessing that the transformer output is a single winding with a center tap.  It feeds a couple of rectifier diodes and a filter cap.  That feeds both of the ICs, one for +12 and one for +5V.

Reverse Converter

120 VAC input <= 44 Watts
240 VAC output <= 40 Watts
Enercell
Radio Shack 273-380

	There is a IEC 60906-1 (Wiki) socket just above the label.  4.5 mm dia x 19 mm C-C. A round pin that's 4.2 mm dia fits nicely in the socket.  The center-to-center distance between the socket pins is 19mm, not the 17 mm Wiki lists for the Europlug.
	On order from McMaster-Carr p/n 1447K13 International Business-Machine Power Cord for Use in Europe 4.7mm dia pins. 18.93mm C-C Plug marked: LP-34, Linetek, 16/250, E76 and and 9 standards organization logos. There's also what looks like a socket where an additional pin could be installed for a ground connection. When plugged into the ReverseConverter the socket in the plug lines up with what appears to be a hole to allow the Reverse Converter to the screwed into the wall socket.

Noise Source

Got this on eBay with title "noise source/Simple spectrum external tracking source" by seller fly-xy.
claimed performance is about the same as actual.
ENR (similar to the output signal strength) : 50KHZ-1.5 GHz, better than 50DB;1.5 G-2G is superior to 40DB, 2G-2.6G is better than that of 30DB.

Network Analyzer

10 Hz to 500 Mhz frequency range with a linear or log sweep.
With just the 4395A there is very limited Network Analyzer functionality.  You can do a simple through type calibration that corrects for amplitude variations but does not do vector error corrections. 

Test Sets

To get really useful results it helps to have some type of test set.  There are a number of options:

• 87512A is a 50 Ohm passive DC to 2 GHz Transmission - Reflection box that will give you S11 and S21 at the same time, but you need to remove and reverse a device to get it's S22 and S12 parameters.  If only the S11 and/or S21 is required or if you need S-Parameters below 300 kHz this is the one to use.  The 87512B is a 75 Ohm version.  These come in a carry case with a short and 50 Ohm termination for doing the calibration.
• 85046A is a 50 Ohm rack width 300 kHz to 3 GHz S-Parameter Test Set that interfaces to the test set control DB25(f) connector on the back of the 4395.  It uses relays to switch from forward to reverse87512 measurements and can measure all 4 S-Parameters without disturbing the device under test. The 85046B is the 75 Ohm version.  You will need a calibration kit and cables & adapters for you device under test.
  K6JCA - Modifying a 75 ohm HP 85046B S-Parameter Test Set to be 50 ohms (2015 - the eBay price of the 8503A + 85046B is about equal to the 85046A)
  
• 87511A is very similar to the 85046A except it is rated for 100 kHz to 500 MHz. The "B" version is for 75 Ohms.
• Other HP/Agilent test sets also will work but may not be cost competitive.
• For a simple way to measure S11 I am trying a Mini Circuits ZFDC-10-1 Coupler.  Spec is 1 to 500 MHz. The connection is:
• RF Out to coupler IN
• coupler CPL to A
• 4395A ports R and B not used.
• keystrokes: Preset, MEAS\analyzer type\network\more\A, CAL\calibrate menu\S11 1-port connect open. short and load then Done
The screen dump of the same 50 Ohm Termination used for the cal (using 16 averages, for both cal and measure) shows pretty good performance from 2.5 MHz on up.  The 4395 needs to warm up for at least 30 minutes, otherwise the calibration will change a lot.  After adding a Splitter (3 each 16 2/3 Ohm resistors in the "Y" configuration): RF Out to Splitter, Splitter to 4395A "R" port and to the input of the Mini Circuits ZFDC-10-1 Coupler CPL output of coupler to 4395A "B" port (it is near edge and closer to me).  For BNC cal standards I am using:
• Open = just air
• Short = a BNC shorting Plug that used to have a chain from my junk box
• Load = Radio Shack 278-270 BNC 50 Ohm Load (selected from a number for 50.0 Ohms at DC)
Prior to cal Averages was turned on (Bw/Avg key) and the default 16 averages was used.  The 4395A had warmed up for many hours.  Just after cal the return loss was around 60 to 70 dB.  The next day the return loss is better than 40 dB.  Note that this is a repeatability test since I am measuring the same load as was used during the calibration.  I have not yet tried making some type of verification standard.  The reference channel really helps!   There are a couple of short BNC cables and a BNC(m) barrel in this setup.  If these cables are wiggled the return loss will degrade.  That's the nice thing about a test set, all the connections are stable.
• 43961A (see below) - This was made for the 4396 combination Network/Spectrum/Impedance analyzer the came before the 4395A and so covers it's frequency range of 100 kHz to 1.8 GHz.  The test port is an APC-7 connector facing up and has two posts to hold a family of 7mm test fixtures.
• 41951A - This is very similar to the 43961A, but was made to work with the 4195 over the 100 kHz to 500 MHz frequency range.  Note that the 4395 was designed to replace the 4195A.  Like the 43961A it holds the 7mm plate type fixtures.
  

4380A

I worked at HP/Agilent in Santa Rosa, CA on the 4380S system for a couple of years.  (see 1998 catalog page 269) This was an HP idea ahead of it's time.  HP pioneered 10-BaseT (Wiki) so that existing phone wiring could be used for computer networking instead of pulling new coax.  Later this was upgraded to 100-BaseT which evolved into CAT5 (Wiki) cabling.  These networking methods use Unshielded Twisted Pair "UTP" (Wiki) and the electrical characteristics of this balanced line are very important.  Prior to the 4380S balanced test system pairs were measured using BALanced to UNbalanced "BALUN" transformers (Wiki).  But as the frequency of LAN cabling increased the unknowns in the balun introduces a lot of error. Also there are no calibration standards for balanced line.  The 4380A test set was part of a system used to characterize balanced LAN cables at and above the CAT5 level.  This system requires either an 8751A,  4395A or 4396B analyzer, a computer running the Rocky Mountain Basic software and the 4380A test set.

The principle of operation is to connect each wire to a test port, so 8 test ports are needed to connect both ends of 2 pairs.  After a full calibration (or by using symmetry the 4380S software has calibration routines that take much less time) all the S-parameters are measured.  The S-parameters can be transformed, first into Z-parameters, to result in either balanced parameters or unbalanced parameters.  This allows characterizing balanced devices, like LAN cables, and also balanced to unbalanced devices, like BALUNS.

The system impedance, which in the past was a hardware option ("A" suffix on the above test sets for 50 Ohms or "B" suffix for 75 Ohms) is a software option with the 4380S system.  Early versions of the 4380S used an HP Rocky Mountain Basic workstation like the 9000 series.  Later versions ran under Rocky Mountain Basic for Windows.  The last version would run with the 4396B, extending the upper frequency limit to 1.8 GHz.

Agilent has discontinued the 4380S system and sold it to a company called ATN Microwave that as part of it's product line has balanced test sets, that are different in that they work much higher in frequency, but have only up to 4 ports.  Aimed more at modern wireless devices than at LAN cable.

The 4380A test set has a built in Open, Short, Load test port to minimize the time needed for an 8 port calibration.  By using a conventional SMA Cal kit on one test port cable end and also using the internal reflection cal standards the cal kit values can be transferred to the internal cal measurements.  Using the built in test port is much much faster than manually connecting OPEN, SHORT and LOAD terminations to the end of eight cables.  But you still need to connect eight cables to the test port.


There are no control buttons on the front panel of the 4380A so you need to control it using the DB-15 connector on the rear panel.  IBasic program (4380MC.txt) that runs inside the 4395A Network Analyzer to control the Source, A receiver, B receiver and termination.  This version uses the soft keys for control and that makes using the soft keys for cal difficult.  A better way would be to use the display/more/disp alloc/Basic status for the program I/O and a PS2 keyboard for the 4380A control.


The 4380S typically used the 3.5 mm (SMA) cal kit and SMA type cables between the test set and the test fixture.  If the test cables have any movement between the time of calibration and measurement then an error may be introduced in the final result.  So, while flex SMA cables are very easy to bend (shape to connect to the text fixture) using old fashioned 0.141 semi-rigid coax and a bending tool results is a more stable test setup.


The insertion loss from the input connector to the selected source port is about 20 dB from 1 kHz up to maybe 2 GHz.  The insertion loss from the source port to either the A (top) or B (bottom) output ports is about 14 dB.  The IL from the input port to Reference port is about 34 dB, the same as the sum of the prior two.

Network Analyzer Cal Kits

The 4395A has parameters for the following calibration kits built into the firmware:

• 7 mm  85031B
• 3.5 mm 85033D
• 50 Ohm N 85032B
• 75 Ohm N 85036B

You may want to use other cal kits and there is a "user" cal kit that can be setup to use any cal kit for which you know the required parameters.  It turns out that all the cal kits have generic data.  The cal kit parameters are not serial number dependent.  The manual for each cal kit has the required parameters printed near the back.  I wrote an I-Basic program that could be run either in an external Rocky Mountain Basic compatible computer or can be run inside the 4395 (or 4396) to load any of the cal kits offered by HP/Agilent that make sense for the 4395A.

Couplers

There are many types of couplers but understanding the simple coax 10 dB type will help in understanding all of them.  When using a coupler as a reflectometer the labels intended for coupler use are no longer valid for this use.  The connections shown below are one way to connect for reflectometer use, but there's another way where you swap the signal generator and the coupled output ports.  Which of these two you use depends on where  you want the higher path loss and where the lower path loss, between the signal generator and the DUT or between the DUT and the coupled output.

A reflectometer application is shown above.  A signal source is connected to the A port.  There's a small fraction of a dB loss to the D port where some Device Under Test (DUT) is connected.  The DUT has some complex reflection coefficient and reflects a signal most of which goes back to port A, but a tenth of the reflected signal is coupled to port C where it can be measured. 
Note that a tenth of the input signal is coupled to port B, which in this case it terminated in a 50 Ohm load.
Note when measuring reflection coefficients, as in the diagram above, the impedance looking out from port A should be very good to prevent interaction of VSWRs.

In a coupling application a signal source might be feed into port D and the coupled sample on port C might be used in a leveling loop or used as a sample of the source signal.
So how the coupler is labeled depends on how it's to be used.

For the case of the ZFDC-10-1+ coupler it's reflectometer connection is:
Signal source to OUT
Sampled reflection out of CPL to input port on network analyzer
Device Under Test on IN

ZFDC-10-1 Coupler Frequency Response

By using a coupler as a test set S11 (aka VSWR, Smith chart, etc.) can be measured.  This can be a full vector error corrected measurement if open, short and load calibration standards are used.

By setting Sweep - Sweep Type - Log Freq. you can see much more at the low end.  The coupler works very well above 50 kHz.  That's much lower than the specification of 1 MHz.

Measuring Cable Length

Connect the cable to the OUT port of the coupler
At the far end of the cable connect a short
Format - Delay
Marker - 51 - M/u (sets the marker to 51 MHz), the frequency is not too critical since the display is almost a straight line
Read the delay in the top right hand corner (19.262 ns, is about 19 feet out and back in air.  this cable is 6 feet of RG-58A/U. 2*6/19.262 = .623vel factor) (1/C =1/186,000 miles per second = 1.0182 Ft/nsec. )

Impedance Analyzer

With just the stock 4395A you can do impedance measurements, see the Z:Transform method.  Option 010 which is the firmware to calibrate, measure and display impedance and a test set like the 100 kHz to 1.8 GHz 43961A.  This unit has an APC-7 connector pointing up and a couple of pins also pointing up.  These are there to accept a wide range of 7mm impedance plate type test fixtures.

You can use the marker on a Smith Chart display to read out impedance values, you just can not get the nice impedance displays that come with the option 010 firmware.  The accuracy of a Smith Chart measurement for impedance purposes is not very good, the 43961A is somewhat better.  

There is a good Impedance vs. frequency graph in app note 346-4 8 Hints for Successful Impedance Measurements (5968-1947E) the shows the areas covered by different types of impedance measuring instruments.

43961A

The 43961A was made for measuring impedance with the 4396A/B combo boxes that cover 100 kHz to 1.8 GHz.   It is a V over I method of measuring Z rather than a reflectometer method.  This is described in the Impedance Measurement Handbook (5950-3000).  This method has more accuracy than 50 Ohm reflectometer network analyzer for impedance magnitudes that are not close to 50 Ohms.

The upward pointing APC-7 connector and the two posts are designed to hold one of a number of test fixtures described in the Accessories Selection Guide For Impedance Measurements (5965-4792E) The 7 mm fixtures start on page 45.  They are:
16092A Spring Clip Fixture - a general purpose fixture for both axial lead and end contact SMT components
16093A Binding Post Fixture - for leaded components
16093B Binding Post Fixture - has two ground lugs with different spacing to the hot binding post
16191A Side Electrode SMT component fixture
16192A Parallel Electrode SMT Fixture - 1 to 20 m long parts to 2 GHz
16194A 200^oHigh Temperature SMT Fixture -
16196A Parallel Electrode SMT Fixture - 0603 parts to 3 GHz
16196B Parallel Electrode SMD Fixture -  0402 parts to 3 GHz
16196C Parallel Electrode SMD Fixture -  0201 parts to 3 GHz
16197A Bottom Electrode SMD Fixture - 0402 to 3225 parts
16094A APC-7(m) Probe - mates to an APC-7 cable
16200B External DC Bias Adapter - to 5A or 40 Volts
16453A Dielectric Test Fixture - uses flat sheet samples
16454A Magnetic Material Test Fixture - must have toroidal sample

When used with the 4395 the low frequency coverage does not go quite all the way down to 10 Hz, but maybe down to 1 kHz using a Network Analyzer S₁₁ calibration.

The 43961A comes in a suitcase that's about 16" x 12" x 4.5" and includes the fixture, a short N(m) to N(m) cable for the source (Note that is the source cable changes due to any motion, that change will be canceled out by the Reference channel.). APC-7 Calibration standards of 50 Ohm Load, Short and Open.  The open is actually the axial lead test fixture that has a 7 mm bore that was used with the 4191A Impedance analyzer.  That's why there is a threaded hole in the top, to hold the axial lead attachment fitting.  It's marked 04191-85302 D: 7 mm, H: 20 mm.  There also was a 10 mm bore axial lead fixture.

There's also a couple of 3.5" floppy disks:
One is marked "HP 4396A Sample Program for External Controller Rev A.01.00 p/n 04396-61002 Disk 1 of 2"
The other is marked "Sample Program Disk for HP Instrument Basic"  Rev. A.01.00 p/n 04396-61002 Disk 2 of 2"

There's a second 3.5" floppy disk wallet with a blank disk and label. The programs on the disk seem to be small example programs, identified by Figure numbers, from the 4396A/B programming manual.

42961A is not directly useable on a 4395 or 4396 that does not have the $1k optional impedance firmware installed.  This because the V (voltage) reading is on the R ( Reference) port and the I (Current) reading is on the A (left) port.  If you make an A/R measurement you get I/V (i.e. admittance).  Also the manner in which the calibration factors are determined and applied for a reflection measurement, like S11, is completely different from how you make a V/I Impedance measurement.

43961A Results with no calibration at 1 MHz (dB @ degrees / mag @ deg)

	Short	Open	50 Ohm Term
A	-6.06  @ 65 0.2477 @ 65	-97.8  @ 141 1.6595E-10 @ 141	-32.3  @ -64.3 5.83E-4 @ -64.3
R	-6.6  @ -116 0.218 @ -116	-0.795  @ -111 0.8327 @ -111	-3.24  @ -112 0.474 @ -112
A/R	-19.69 @ -178	-97  @ -113	-29.1  @ 177
|R/A|*	0.88 near zero	5E9 near infinity	813 not too near 50?

Since the A port is the current through the DUT it is minimum with an open test port and maximum with a shorted test port.  Note this a lot different from a reflectometer that has the same max. magnitude with either an open or short and is minimized with a matched load.

Changing the test port value has a little effect on the R port value, which in a perfect world would not happen.  But if the value of the R port reading is really the voltage at the DUT, then this is a good thing in that it results in a more accurate impedance measurement.  This very well may be the case since there are three small coax cables going up into the APC-7 test port connector, one of them being for the R port.

* calculated without regard to the phase angles.

Looking at just the A port:
with a short as the DUT, the response rolls off (down) from -20 dB as you go down in frequency below 20 kHz at 12 dB per octave.
with an open as the DUT the response rolls off (up) from -120 dB as you go up in frequency above 1 MHz.  This means that for frequencies in the 20 kHz to 1 MHz area the dynamic range of the 43961A is about  85 dB and it's about 30 dB at 2 kHz and 100 Mhz.  This probably has some bearing on the accuracy.  Interesting that 1 MHz, a common frequency on LCR meters is in the sweet area of the 43961A, allowing correlation at the same frequency.

If you have the 4395-010 Impedance software contact me
16 July 2003 - I have figured out how to calibrate the V/I test set.  Just to confirm some things I connected the 43961A to the 4395A upside down.  This connects the 43961A "R" or V signal to the 4395A A port and connects the 43961A "A" or I signal on the 4395A R port, now when the 4395A is set to display "A/R" the display is LOG(|Z|) and phase(Z).  Using this setup I have manually confirmed my calibration math.
17 July '03 -Bummer, IBasic does not support the COMPLEX file type that normal RMB uses.  Can maybe just do a magnitude only correction, or do the Vector correction in LabVIEW, or with a lot of effort do a vector correction in IBasic?

Z Transform Impedance Method

The metal box connects the outer conductor and the center conductors are brought out to binding posts.  Short the binding posts to each other when doing the response THRU cal.

Background

With a stock, no options, 4395A you can measure impedance without any test sets over the full frequency range of the Network Analyzer (10 Hz to 500 MHz). The result can be displayed as Log Mag, Lin Mag, Phase, Real or Imaginary.

One of the key features of the 4395A is it's low frequency coverage.  The overall frequency coverage is about 8 decades, that's very wide.  The 43961A was made for the 4396A and so does not work very well below 100 kHz which eliminates VLF and VF testing.

Before I discovered this method I was using the same connection and measuring S21 and then using a spreadsheet to convert from S21 into impedance.  Now I let the 4395A do that math.

There is an option under Meas/Conversion for displaying Z:TRANS,  the impedance when the DUT is inserted in series with a transmission path, like an S21 network analyzer measurement.  The Lin Mag can be displayed and reads out directly in Ohms.  If the Log Mag display is selected you can convert back to Ohms using Ohms = 10^(dB/20).

E5100

The E5100() Network Analyzer uses this Z:TRANS connection method as well as the E4915 crystal impedance & E4916 crystal impedance/LCR meters.  The Network analyzer method that is covered in 8 hints for Impedance measurements, pub # 5968-1947E, is the reflectometer method.  But it also mentions transmission as having the same type of errors.  So making up a circuit similar to the 40 dB attenuator above that transforms 50 Ohms up to some higher value (or to some lower value) that's closer to the |Z| you are measuring, the more accurate the result will be.

The IEC-444 Specification (PI-Network Measurement of the Load Resonance Frequency without Load Capacitor.pdf) - From the crystal point of view the PI-Netowrk looks like 12.5 Ohms at both the input and output, or 25 Ohms external loop resistance.  The network also includes shunt capacitors to trim the series phase to zero. (Corning: Technical Introduction Crystals.pdf, PI-net: pdf pg 14) 

The HP 41900A Pi Network Test Fixture also included a number of load capacitors. IEC444
The HP 41901A SMD PI Network Test Fixture also included a number of load capacitors. QIAJ (Japan)
The E4916A  Options:
001 - Impedance Probe Kit
010 - LCR Measurements - eBay has titles that include Impedance/LCR, but the unit may not have option 010?
adds LCR probe HP 16099A capability.
E4916U - upgrade kit adds both 001 & 010 - OBSOLETE

How it Works

The Device Under Test (DUT) is connected in series with the center conductor between the Source and "R" receiver.  The shield is continuous from the source to receiver.  Calibration can be as simple as a response/thru cal.  Setup, where [hard key] and (soft key):

• Measurement area [Meas], (analyzer type) (network analyzer)
• Measurement area [Meas], (CONVERSION), (Z:Trans)
• Sweep area [Sweep], (SWEEP TYPE), (LOG FREQ) - this makes most LF and VF plots easier to understand.
  
• Measurement area [Format], choose from (LOG MAG), (PHASE), or (MORE) then (REAL), (IMAGINARY).
  

  Note that when using LOG MAG you can convert dB back to Ohms using:
  Ohms = 10 ^ ( dB / 20 )
  It's convenient to set the Measurement area, [Scale Ref], {REFERENCE VALUE) to zero (i.e. 1 Ohm) and the (SCALE/DIV) to 10 dB (every 2 divisions is 10 X impedance, or 20 dB (every division is 10X impedance.
  

The 4395A does the transformation math so the display is the device impedance since it knows how the device is connected.  I have seen strange things when using the "A" or "B" receivers and if a power splitter is used.  The best results are with a simple source to "R" connection.

Calibration

Before doing the CAL you may want to change to Measure, Network, R (you can also use A or B and if you have a splitter then use A/R or B/R)
Also change to Sweep, Sweep Type Menu, LOG FREQ.

The RBW (Measurement, Bw/Ave, IF BW) should ideally be 10 times smaller than the start frequency.  The best you can do is 1 Hz RBW.
If the RBW is equal or larger than the START frequency the trace will have a lot of noise at the low frequency end.

The very simple Measurement area, [CAL], (CALIBRATE MENU),  (RESPONSE) can be used. 
With the fixture shorted use only the THRU option then DONE.
Typical values after calibration are:  44 milli Ohms for the short and 1.2 Meg  Ohms for the open.

If the fixture capacitance is a problem then use . . .  (RESPONSE & ISOL'N) with the (THRU) option when the fixture is shorted and after the short is removed use the (ISOL'N STD).
Typical values after the calibration are:  117 milli Ohms for the short & 13 Meg Ohms when open.

Verification

If a resistor is connected and Measurement, Format is set to MORE, Real then and MARKER pressed the value should be the value of the resistor (allowing for it's tolerance).

When showing LOG Frequency and Measurement, Format LOG Mag a capacitor will show as a straight line sloping down to the right with a slope of -20 dB per decade of frequency. 

ignore the 2 preceding the Z it's a typo.

To convert log MAG into Ohms use:  Ohms = 10 ^ (dB/20) For example at 2 kHz 120.52 dB = 1,061,696 Ohms Note: for a Response cal the open impedance is just over 1 M Ohm, see above. and at 20 MHz 48.958 dB = 280.5 Ohms. Note this is combined real and imaginary so may not be equal to the imaginary values shows at the right. The frequency range below about 20 kHz is noisy because of the RBW being close to the test frequency instead of being 10 or more times smaller.

The calculated capacitive reactance is: 2 kHz:   -3,180,000j 20 kHz:    -318,000j 200 kHz:    -31,800j 2 MHz:        -3,183j 20 MHz:         -318j Using -285.5 Ohms the capacitance is (1/Xc)/(2{PIf) = 27.87 pF Cap marked 27 in bag marked 25.

An inductor will slope up to the right at +20 dB per decade of frequency.

Z:Trans Fixtures

No Bias Fixture

This is a simple Ponoma metal box with a couple of BNC connectors.  The metal box connects the shields and the center conductors are brought out to a couple of binding posts.  The separation of the binding posts needs to be close enough to connect the DUT but as far apart as possible to minimize leakage capacitance. 

This fixture has about 65 pF series capacitance.

Another fixture could be made for measuring crystals that would have the standard PI network installed.  If the impedance looking back and the source and into the "R" receiver from the DUT is other than 50 Ohms then you need to change the Zo value at Measurement area, [CAL], (MORE), (SET Z0). 

Bias-T Fixture

This fixture was made using 2 Radio Shack Filament transformer primary windings for the 2 inductors and the two large DC blocking caps were made by paralleling 15 each 2.2 uF 250 Volt Mylar caps.  Each of the BNC connectors that go to the 4395A have a pair of LEDs connected from the center conductor to ground.  That way you can experiment with the fixture disconnected from the 4395A to see what lights the LEDs and AVOID doing that during the testing.  The danger is that the caps or inductors will get charged and dump their energy into the 4395A letting the smoke out and giving you gray hair.
See the model 500 dial phone tests on my Zo web page for more on using this fixture.

This fixture was made for testing in the Voice Frequency range, but a similar fixture could be made that covered the full range of the 4395A.

10X Scope Probe

It occurs to me that you could use a 10X Scope Probe.  The probe is designed to divide the input voltage by 10 when the probe is connected to a scope with 1 M Ohm input resistance.  But when connected to a 50 Ohm input the division ratio is in the area of 20,000 times or about 43 dB.   The dynamic range of the 4395A can handle that for many applications so this may be a good way to get DC voltage attenuation and/or a high impedance input.  The frequency compensation will be wrong, but with such a low load resistance should still be very good.

Other Impedance Info

HP 4191 & 4193 patent 5345182 Impedance meter capable of performing measurements at high precision over wide impedance and frequency ranges Sep 6, 1994 324/649; 324/522; 324/654; 324/713

This is the patent that covers the V/I type impedance measurements pioneered in the HP 4191 and 4193 instruments and later used on others.  It works at frequencies higher than can covered with the very high accuracy Four Terminal Pair method.

Save State

There are a lot of key presses to get the Z:Trans method working.  Once you are happy with a setup it makes life much easier to then save the instrument state by being sure the: Instrument State, [Save}, (STOR DEV) is set to DISK, then in the Instrument State area, [Save], (STATE).  If you connect a standard PC keyboard to the 4395A (cannot use PC keyboard on the older 4396A) it's much easier to enter the file name and you get upper as well as lower case letters.

By using a PS2 Keyboard & VGA Monitor (Wiki) switch  you can just press a button to switch your keyboard and monitor back and forth between your computer and the 4395A.  This makes looking at the display much easier and larger.  This was true in the day of PS2 keyboards (Wiki) but no longer the case with USB keyboards and mice.

To get the 4395A back into that state be sure the floppy disk is installed, then in the instrument State area, [Recall] which after a few seconds of reading the disk directory will display all the available states as softkey choices and you just press the one you want.  If you have saved IBasic programs from the 4395A to floppy they'll also show up as softkey choices.

Signal Generator

In the N.A. mode you can set the Sweep\Source Power from -50 dBm to +15 dBm and the CW frequency from 10 Hz to 510 MHz (yes a little more than 500 MHz).  Instead of using the CW frequency you can set the start and stop frequencies and number of points to generate a ramp that will show up as an FM modulated signal that will pass through the audio output of a receiver.

In S.A. mode with SPAN set to zero you can set the signal generator to the same ranges.

Note that in either case the signal generator is coming from the respective function (N.A. signal source or S.A. local oscillator), so the mode will have limited or no use when the signal generator is being controlled directly.

Floppy Disk Drive

The drive will accept today's standard double sided high density (1.44 MB) type disk used on a PC.  When you Initialize (format) the disk using FORMAT [DOS] and STOR DEV [DISK] the disk will be just like a normal PC 1.44  floppy and the title will be HP4395 (even if the 4395A has an Agilent logo).  The disk can be read on a PC.  But the HP LIF format can not be read on a PC.
I don't think pre-formatted HD floppy disks will work, but they may if the 4395 formats them as standard density?

Initialize Floppy

In the Instrument State box press the [SAVE] button.
Choose the on screen menu item: File Utilities, then Initialize Disk, then toggle Format[DOS] so that DOS is selected, not [LIF].  This way a PC can read the disk.
Then select Init Disk[Yes].

Display, Keyboard & Mouse

The screen video and a keyboard can be connected to rear panel jacks.  By using a PC switch box these can be connected to the VGA monitor and PS2 keyboard associated with your computer.  The computer monitor is a much larger screen and the keyboard is great for entering file names when saving screen shots.  This box is also connected to two computers and has one free input port.  A good thing would be to find a scope that can put it's display on the monitor.

Time Gating

By using time gating only signals that occur in the allowed time window will be analyzed.  When testing hard disk drives with the E5022A or E5023A spin stand systems you can restrict the analysis to only a specified sector of the disk.  There was a version of the 4395A spectrum analyzer (only) made in VXI format  named the E5040A used in the E5023A system.

Applications

Audio

Checking a dynamic microphone on the H-250 Handset

Remember that the bandwidth needs to be much narrower than the lowest test frequency.  When the default BW of some kHz is used with a lower frequency like 20 Hz the lower part of the display has a lot of ripple that will go away when the BW is set to 10 or 3 Hz.

Amplifiers

By using an antenna to monitor a transmitted signal and switching an RF amplifier in and out of the transmitter circuit you can see the gain of the amplifier.  This avoids the need for high power attenuators which would be needed for a direct connection.
OF-185 Amplifier gain -

Antennas

In spectrum analyzer mode you can compare different antennas to see which gives stronger signals
OE-254 vs. Radio Shack
T2FD vs. DA-100 @ 15 MHz note the difference in noise level and signal strength.

In Network Analyzer mode you can measure the Smith Chart.
OE-254 Smith Chart

The AMRAD LF active whip antenna was found to have a noise floor problem that was very easy to see using the 4395A.  Before and after modification plots are on the web page for the antenna.  This type of problem would be very hard to detect any other way.

An ideal receiving antenna should have good signal to noise ratio for the desired signal and reject unwanted signals.  What "good" means depends on what's going on.  For example for a TV signal to look excellent and without any trace of noise you need about 48 dB s/n.  Some slow speed digital modulation schemes will recover a an error free copy of the transmitted signal with a s/n of maybe -20 dB (note the minus sign).

A spectrum analyzer is a great tool to look at receiving antenna s/n ratios.  The network analyzer is great for measuring antenna properties.  12 May 203 working on Ferrite WWVB 60 kHz receiving loop antenna.

The inductance can be calculated from the series resonate frequency in an S21 measurement, but the HP 4332 LCR meter goes nuts because the loopstick is picking up strong signals from the computer monitor and who knows what.  The insertion loss at series resonance allows the effective AC resistance of the coil to be easily calculated. For 50 Ohms source and load (like the 4395):

R = 100 * [ 1/ (10^(dB/20)) -1]
Example for -1.0 dB R = 12.2 Ohms

Note that in a vector network analyzer the receiver is phase locked to the source and so external signals have minimal to no effect.

Crystal Oscillators

Many radios use crystals to set the receive and/or transmit frequency.  By using the 4395A in spectrum analyzer mode and connecting the radio antenna to the analyzer, you can see most of the crystal frequencies either at the fundamental or some harmonics. CAUTION - Do not transmit into the 4395A without making some provision to reduce the power so as to not smoke the 4395A.  For transmitter testing I connect the 4395A to an outside antenna and use the hand held radio with a short antenna.  The 4395A is acting like a receiver.
Finding the Crystal frequencies and the formula for the URC-68 radio.
Confirming that crystals are active in a PRC-68A radio.

Crystal Equivalent Circuit

By connecting a crystal in series and measuring the S21 mag and phase in Network Analyzer mode a computer program should be able to curve fit and come up with an equivalent circuit since the circuit is already known only the values would need to be fit not the topology.

Phase Noise

One of the ways of making a phase noise measurement is to just look at the signal on a spectrum analyzer.  With the older analog non synthesized analyzers this did not work very well, but with a synthesized analyzer you can get useful information although there are corrections for the non square RBW analog filters and for the peak detector used in analog units.  But the 4395A has very steep skirts on the digital filters and has a true RMS detection mode that works by squaring the I and Q signals.  When these options are selected using the on screen menu that display is directly in dB per root Hz.
User-Friendly Phase-Noise Measurement for HP and Tektronix Spectrum Analyzers - this free software may work with the 4395A.  haven't tested it.

Spectrum Plots

0 to 200 kHz
200 to 400 kHz
40 to 50 MHz 3 Hz RBW, Max hold, many days
118 to 136 MHz
136 to 138 MHz
150 to 154 MHz
153 to 157 MHz

150 MHz center freq, 1 Hz RBW, Max hold, 24 hours - looking for satellites and found 4 of them. (17 Apr 2013).
The Doppler (Wiki) effect causes the frequency of the satellite to change.  The width shown on the spectrum plot is determined by the satellite elevation.

400 MHz center freq, 1 Hz RBW, Max hold, 24 hours - looking for satellites and found . . . 1

450 MHz center freq, 1 Hz RBW, Max hold, 24 hours - looking for satellites and found . . . 2

VCO Testing

When the option 001 (A2 PCB) Programmable DC supply is installed an output of up to +/-40.001 V < 100.02 mA of DC voltage/current can be programmed.
While the 04395-90010 (page 319) does not mention noise, AFAICR it's very low to allow testing Voltage Controlled Oscillators (Wiki).
A better way is to use the HP 4352B VCO test system.

E5040A VXI Spectrum Analyzer

This is the Spectrum Analyzer part of the 4395A. There is no impedance or network analyzer capability, only spectrum.  It was part of the E5023 Hard Disk Test System.
This is a Series C size VXI module and needs either a computer in the VXI cage or an adapter to a computer in slot zero.  It does have the option for time gate in order to measure various sectors of a hard disk.  The Gate out can be fed to an oscilloscope so that you can adjust it via software to cover the sector desired. It also has external reference in and out to frequency lock with the test generator.

The manuals for the E5023 system do not cover the E5040A. Just use the 4395A commands applicable to spectrum analysis with the E5040A.

Keysight: VXI Mainframes & Controllers -
E1406A Command Module C-Size
E8491B IEEE-1394 PC Link to VXI C-size
E8401A C-Size 13-Slot Mainframe
E8403A C-Size 13-Slot Mainframe
E8408A C-Size 4-Slot Mainframe - like the small size.
E8491B Firewire Controller C-Size - FW replaced by USB2 so not a good choice.
E1406A Command Module C-Size - GPIB, RS-232, Slot 0, and Resource Manager - link to computer not one itself - Docs available
E4220B Transportable VXI 7 Slot Housing - Monitor, Keyboard & Trackball, no documents (the A version also has no documents)
V743 C-Size Controller based on the 9000 Series HTB computers. uses PS/2 (Wiki) keyboard and mouse & SCSI hard drive interface
The E1421A C-Size 6-Slot Mainframe has been replaced by the E1421B with many support documents
eBay search terms: "(HP,Agilent,Keysight) VXI Mainframe"  Note B-Size is the same as B-Series and smaller than this C-Size spectrum analyzer.

Fig 1	Fig 2
Fig 3

Reference Information

The 4191A was a precursor impedance analyzer to the 4395A.  The 4195 was the direct precursor.  Here are some YouTube videos about the 4191A:

4191A part 1 an introduction - Visual Basic for Excel, Simple Example Program to Control Instruments,  Jan 1980 HP Journal,
4191A part 2 the software -  Based on Excel
4191A part 3 calibration - short, open , load
4191A Part 4 making a measurement using Z plots - Link to ZPlots with thanks to Dan Maguire

Note most Agilent Plug & Play instrument drivers include LabVIEW capability.

Firmware

Re: HP4395A - options upgrade
From: Hristo the complete method
4395A_OptionsUpgrade.pdf

Scalar Analyzers

Related

ADSB &Raspberry Pi
Agilent E4404B ESA-E 9kHz - 6.7 GHz Spectrum Analyzer
Avcom PSA-45D Portable Spectrum Analyzer
HP 71100C 2.9 GHz Spectrum Analyzer
HP 8702B Lightwave Component Analyzer, Electro Optical Network Analyzer -
Optical Spectrum Analyzers
Test Equipment - Microwave Test Equipment - Spectrum Analyzers & some 70000 info
Z - Impedance
Zo - Transmission Line Zo vs. Frequency

Links

Back to Brooke's Test Equipment, Microwave Test Equipment,  Electronics, Home page

Page created 28 Feb 2001.